I worked on many many medically related projects where it was necessary to publish flow charts for all the implememted code whether that be stuff writtn in C, assembler or otherwise.

My experience is that trying to use a general tool to generate flow charts from the code itself gives you a pile of rubbish as a result. The results are almost as bad as the programmer that adds comments to the code that repeat exactly what the code is doing such as:

     MOX      A, B        ; move B to A

Flow charts should move up a level from the code and show flow and decision trees and be annotated in language that a non-programmer can understand. I find it best to just open up Visio and start drawing flow charts of your code. If done corectly this can be a very valuable exercise. First off it imposes a type of extremely detailed self design review of the code. Secondarily it will also reveal weaknesses in the code flow or expose outright bugs that need to be fixed.

Another thing to consider is that many times when you make these flow charts they are rarely reviewed by anyone unless some later problem springs up and somebody goes searching for where to point the finger.

The only time I have seen successful auto flowcharting work at all was when the developers were required to document the code in the source files in a way that included adding special tags and category titles and texts. Then a propritary in-house tool is used to parse the source files to pull out these tags and comments where they are merged with an analysis of the code flow to form charts. Only problem I had with that was that the resulting charts were next to useless becasue they ended up being these long linearized flows that ended up being printed out on long stacks of paper and it was next to impossible to gain any referential context of one part of the code versus another like you would get from a carefully designed flow chart.

Michael Karas